Two recent well-designed studies of tricyclic antidepressants (TCAs) for children with major depressive disorder (MDD) have had negative findings. We are proposing that the reason, in part, for this lack of effectiveness of TCAs in pediatric MDD is that the subjects were heterogeneous with respect to a future course of unipolar or bipolar disorder. Further, we are hypothesizing that children with MDD who also have predictors of a future course of bipolar disorder will be responsive to treatment with lithium. Based on the above, we plan a three year study of 60 prepubertal subjects (30 active, 30 placebo) 6-12 years old who have DSM-III-R MDD and either of the following two categories of predictors of a future course of bipolarity: (1) a family history of bipolar disorder; (2) a loaded or multigenerational family pedigree exclusive of a bipolar family history. (i.e., Subjects who have both a bipolar family history and a loaded or multigenerational family pedigree will be counted in the bipolar family history group). Treatment groups (active vs. placebo) will be balanced for these two categories. An eight week, random assignment, double-blind placebo-controlled design is planned. This would include a two week single-blind placebo-washout phase and a six week double-blind placebo-controlled phase (two weeks to adjust dose and serum lithium levels and a minimum of four weeks within the protocol serum lithium range). Assessments would be performed utilizing the pediatric versions of instruments developed for use in studies of adults and include psychiatric, psychosocial, family history, cognitive, side effects and physical measures. Renal, thyroid and parathyroid function; CBC, ESR and SMA-18 levels; and EKG, height and weight measurements will be obtained pre and post protocol. Current data suggest that MDD in children is a chronic disorder which seriously interferes with age appropriate family, peer and school functioning. Since there is a lack of proven treatment, these children are likely to have continued morbidity and suicidality into their adolescent years. There is, therefore, a crucial need to develop effective interventions.